MacInTouch Reader Reports

Hard Drives: SSD & Hybrid

Feb. 10, 2009
Feb. 17, 2009
Feb. 18, 2009
Feb. 19, 2009
Feb. 20, 2009
Feb. 21, 2009
Feb. 23, 2009
Feb. 24, 2009
Feb. 25, 2009
Mar. 16, 2009
Mar. 17, 2009
Mar. 19, 2009
Mar. 20, 2009
Mar. 21, 2009
Mar. 23, 2009
Mar. 24, 2009
Apr. 8, 2009
Apr. 9, 2009
Apr. 10, 2009
Apr. 29, 2009
May. 7, 2009
May. 21, 2009
Jul. 21, 2009
Jul. 22, 2009
Jul. 23, 2009
Jul. 24, 2009
Jul. 25, 2009
Jul. 27, 2009
Jul. 28, 2009
Jul. 29, 2009
Aug. 4, 2009
Mar. 10, 2010
Mar. 11, 2010
Mar. 18, 2010
Apr. 26, 2010
Apr. 27, 2010
Apr. 28, 2010

Newer entries...
Feb. 10, 2009

item.86910

Robin Lake

'twas asked:

"Related to this, has anyone installed an SSD into the Mac Pro? A physical adapter is needed--the IcyDock ($22 on NewEgg) looks like it might work. Any one have advice?
I am thinking about a boot SSD with /var, /tmp on a scratch drive (or another SSD) and /Users on my 2x1 RAID-1. "

I designed and developed one of the first SSDs, the Extended Memory Unit produced by Monolithic Systems Inc. (now defunct) in the 1970's. Expectations for the performance of the device as a system drive was considerably less than anticipated, simply because the disk device driver software was written for a drive with 17,000 milliseconds latency --- not the 1 microsecond latency of the SSD.

Shreekumar Shenoy did his doctoral dissertation on SSD performance at Case Western Reserve University in the late 1970's.

item.86971

Ted Selvaggio

Check out BareFeats.com, they tested SSD's in MacBook Pros and Mac Pros.

Specifically:
  http://www.barefeats.com/hard103.html
where they stripped two Intel X25-M Solid State SATA 80GB Drives and 'they go like "stink."'

Included are benchmarks and hardware needs.

Feb. 17, 2009

item.87247

Ric Ford [MacInTouch]

Wozniak solid state storage scientist at Fusion-io

Prior to his appointment as chief scientist at Fusion-io, Wozniak was a member of the company's advisory board, where he counselled the company on market trends, product road maps and other strategic activities. Wozniak will continue to advise the Fusion-io team on these vital issues.

The company's ioDrive is claimed to be the first direct-attached, solid-state storage technology on a PCIExpress (PCIe), with an I/O performance that is claimed to surpass that of mechanical disks by hundreds of times.

Fusion-io is currently working with IBM on its Project Quicksilver to easily achieve over 1,000,000 IOPS by presenting multiple ioDrives as a shared storage solution.

Feb. 18, 2009

item.87262

Paul Ukena

The iodrive only works on Windows... thanks, Woz.

item.87303

MacInTouch Reader

Wonder how much time Woz will be spending on SSD and how much on Dancing With The Stars? ;-)

Feb. 19, 2009

item.87326

Lyman Taylor

Re:

The iodrive only works on Windows

No. It works on Linux also. It wouldn't be surprising if those were the bulk of the current installations.

The "drive" requires a unique device driver. At present the Linux server market is far, far bigger than the Mac OS X server market. Apple seems to be taking longer and longer to update the Xserve (there is only full sized PCI-e slot). They nuked the XRaid on short notice. As a small company with limited resources which server markets would you focus on first: Linux, Windows, or Mac OS X ?

(I guess someone could put a couple in a Mac Pro, but the stereotypical I/O hogs there are stuff like video processing which tend to do bulk sequential writes (as well as reads). Sequential writes is where this solution is weakest. Random I/O (e.g., like databases or some application with large number of concurrent users.) is where the huge $/performance win is. How many folks first reach for Apple hardware to run high throughput databases?)

item.87353

MacInTouch Reader

Paul Ukena wrote:

The iodrive only works on Windows... thanks, Woz.

Perhaps Woz took the job to expand the market to Macs. The 2008 MacPro has 2 PCI Express x4 slots--one used for video card.

I don't know of a reason why the second slot, if not being used for a second video card, could not be used if the Mac firmware supports booting from a drive connected to a PCI slot. My G4 was able to boot from a SATA drive connected to a PCI slot.

The Fusion-io card may need a firmware update to work but the MacPro is x86 and 64-bit.

For that matter, is there any reason why an SSD SATA drive can't be connected to a SATA bay or port? It's an option in MacBooks. Obviously it may be more expensive for a while.

Feb. 20, 2009

item.87378

Brian Rafter

You can easily mount an SSD in the Mac Pro's SATA bays. I've got a pair of 80GB Intel SSDs striped that I'm using as a boot drive. I get up to 160 MB/sec for sequential writes and 420 MB/sec for sequential reads. These are great numbers for two drives, although better write performance would be nice. Random I/O measured in transactions per second is an order of magnitude faster than traditional drives, making a good SSD an excellent choice for a boot volume. As a bonus they're whisper quiet, run cool, and have no moving parts. Highly recommended.

Note that if you want to use a 2.5 inch drive, SSD or otherwise, in the Mac Pro's bays, you'll need an adapter such as SZ-VRMACPRO-01 from MaxUpgrades.com (no affiliation, just a satisfied customer).

Feb. 21, 2009

item.87444

Ed Sikorski

After reading Brian Rafter's trial with RAID SSDs (I suspect X-25M models since faster-pricier Extreme SSDs are 64GB models) I found an interesting degrading performance with SSDs article. Seems over time, the performance degrades far greater than standard hard drives:

http://news.cnet.com/8301-13924_3-10168084-64.html

item.87460

Julian Welch

Another option for running 2.5" drives in a Mac Pro is TransIntl's Pro Drive. It a carrier that slots into a Mac Pro drive bay, holds two 2.5 inch SATA drives and can be switched to either RAID 0 or 1. A bit pricey at $169, but a great idea.

Feb. 23, 2009

item.87503

David Solberg

One site (PC Perspective) has found a problem with the Intel X-25M drive speed in long-term use, and every other site reported that problem. In reality, it's just the experiences of two people, not a widely-observed problem. That's not to say that there might not be an issue with long-term use of the X-25M drives, but it's far too early to draw any conclusions.

For its part, Intel has attributed the problem to the testers. I don't know the answer, but I do know that a couple of techies with lots of time and software settings to fiddle with can manage to screw up a working computer pretty easily.

I'm on the fence about getting one of these, and I'm probably going to wait for confirmation one way or another before making a decision.

item.87525

MacInTouch Reader

Solid state disc drives (SSDs) have finite lifetimes as is true with any discrete or integrated circuit device. Many have experienced IC failures in DRAM or microprocessors.

I work in the semiconductor industry and "solid state" is a bit misleading. You do not have an electromechanical product like a conventional hard disk drive (HDD) with visibly moving parts but that does not mean lack of movement at the molecular and atomic levels.

Within any semiconductor device, there is considerable electron and hole mobility during operation. Since transistor junctions are not 100% efficient, heat is generated which will alone contribute to device failure over time.

Do I use solid state memory? Yes?

Do I back up the data? Yes.

SSDs certainly will get better over time but solid state devices all fail at some point.

SSD have many advantages over HDDs but they are just one step ahead of HDDs. Completely different storage computing storage and processing devices are being developed now.

If you store data on any storage media, HDD, SSD or what ever, back it up--preferably on more than one type of media.

Feb. 24, 2009

item.87554

Nick Batzdorf

An unspecified MacInTouch reader wrote:

"Solid state disc drives (SSDs) have finite lifetimes as is true with any discrete or integrated circuit device. Completely different storage computing storage and processing devices are being developed now."

Could you possibly elaborate? How would you compare an SSD's expected lifespan to a conventional HD? And what's coming?

SSD looks like a very interesting technology for the application I work with - running large musical sample libraries (in which fast reading but not writing is important) - but at the moment it doesn't quite seem to be there. Is that expected to change in the near future?

item.87561

David Charlap

A MacInTouch Reader wrote:

"Solid state disc drives (SSDs) have finite lifetimes ... If you store data on any storage media, HDD, SSD or what ever, back it up--preferably on more than one type of media."

Both true, but one statement doesn't depend on the other. Even if you you had a theoretical storage medium that was 100% reliable, you would still be well-advised to make regular backups.

Data can get lost for many reasons other than drive failure. Buggy or malicious software could corrupt files (or the file system), the computer could be damaged by a power surge or a fire or a flood or other disaster. Or you might simply delete the wrong file by mistake (or on purpose, and have second thoughts a month later.)

A backup (especially one stored off-line) is your best protection against all these kinds of data-loss situations.

item.87568

Lyman Taylor

SSD drives that are packaged as SATA drives are easy to install as a SATA drive. (after all that is one of benefits of packaging it as a SATA drive). However, not all SSD drives have SATA drive packaging. Nor do they necessarily use Flash memory. Some come with Fibre Channel interfaces and others, like fusion-io, effectively come with a PCI/PCI-e interface.

The downside of packaging as a SATA drive is that read performance is going to get capped at about SATA 3 Gb/s levels (3 Gb/s is about 375 MB/s. or worse if SATA 1 interface 1.5 Gb/s.) 420 MB/s is only 115% of 375. That's significantly short of a 2x multiplier. The problem is that Flash reads are substantially faster than that. So the read performance is throttled by the interface. So for install convenience take the hit on performance; it is a tradeoff.

As long as SAS and SATA are the only standard drive interfaces and are constrained by rotational drive throughput, will take a hit on performance.



Feb. 25, 2009

item.87636

MacInTouch Reader

Re my initial comments on

"Solid state disc drives (SSDs) have finite lifetimes as is true with any discrete or integrated circuit device. Completely different storage computing storage and processing devices are being developed now."

Above absolute zero temperature, everything "solid" has movement, from Physics 101. Movement of atoms can cause problems.

As to SSD lifetimes versus electromechanical HDDs, it "generally" should be ten times or more longer than HDDs. It is probably much longer but the devices have not been used in volume long enough to know yet. New "solid state" devices in R&D now should last for decades.

One problem, an area I work with, is the mass production microelectronics manufacturing process itself. No chip maker has 100% good yields on every device. Even with devices that appear to work, there is the "infant mortality" factor. Flaws or impurities in the materials used to make SSDs or ICs can, and do, lead to failures.

Personally, since infant mortality tends to occur within the first 72 hours of operation, for Macs or any other electronics I purchase, I leave them on for at least 72 hours. Then after a follow-up week of use where the product is turned off each day and restarted the next, gross thermal cycling shock failures will show up. If, at the end of 30 days, nothing has failed, I'm reasonably confident that the product will work for a few years BUT I BACK UP EVERYTHING.

So what makes sense for the next generation of nanoelectronics devices including microprocessors and memory? To me, an intelligent self-monitoring and self-healing device is the best solution.

As David Charlap wrote,

"Even if you you had a theoretical storage medium that was 100% reliable, you would still be well-advised to make regular backups.
Data can get lost for many reasons other than drive failure. Buggy or malicious software could corrupt files (or the file system), the computer could be damaged by a power surge or a fire or a flood or other disaster. Or you might simply delete the wrong file by mistake (or on purpose, and have second thoughts a month later.)
A backup (especially one stored off-line) is your best protection against all these kinds of data-loss situations."

David's advice is wise since component failure is just one possibility. Ever updated your system or an app and had data corruption or loss problems? Most of us have had this problem with components that were working fine.

One exciting device in the works by IBM's Almaden Research Center in San Jose, CA., is Spintronics "Racetrack Memory" which promises high performance and capacity, low cost and power use. Other companies are pursuing similar research. IBM has accelerated commercial introduction of these device from 2018 to 2011 according to the lead researcher with whom I discussed the device's potential.

Racetrack Memory is based upon IBM's R&D with nanomaterials and nanodevices using the manipulation of spin-polarized electrical current. [See IBM, ITRI Collaborate to Advance New Solid-State Memory for details.]

Racetrack Memory research could lead to electronic devices capable of storing far more data in the same amount of space than is possible today, with lightning-fast boot times, far lower cost and unprecedented stability and durability.

Racetrack Memory is named because the data "races" around a nanowire "track," which could lead to solid state electronic devices capable of holding far more data in the same amount of space than is possible today. This technology could enable a handheld device such as an mp3 player to store around 500,000 songs or around 3,500 movies " 100 times more than is possible today " with far lower cost and power consumption.

The devices would not only store vastly more information in the same space, but also require much less power and generate much less heat, and be practically unbreakable; the result: massive amounts of personal storage that could run on a single battery for weeks at a time and last for decades.

Bottom line: Semiconductor devices including SSDs get better with each generation. I currently use two 16 GB flash drives daily to back up everything as I'm working in case something happens. I also backup using SuperDuper! to three external HDDS on a rotating daily schedule.

No, I do not work for IBM or own any IBM stock. It would be a personal conflict of interest. IBM is the leader in patents and does many things extremely well. I wish them and other researchers well.

Mar. 16, 2009

item.88967

David Foster

I have no info on Apple's outfitting, but I installed a 250GB OCZ Vertex SSD drive in my 17" [MacBook Pro] and it is wicked fast. I suspect that the Vertex drive is probably significantly faster than the one that Apple installs.

Mar. 17, 2009

item.89081

Dafydd Williams

I've had similar success with OCZ Core v2. SSDs. I've put on in my first generation Core Duo Intel mini, and one in a mid 2007 Core 2 Duo Black MacBook.

Both devices now have boot times from cold to fully loaded Finder that are under 8 seconds. Both launch every application in the dock faster than I can click them, all at the same time.

In fact, they're so fast to boot that I don't even put the MacBook to sleep any more; I just shut it down.

Mar. 19, 2009

item.89181

Doug Edwards

Has anybody tried the "Patriot Extreme Flash, 256GB Warp SSD Drive 2.5 SATA V.3" drive or knows how it compares with the OCZ Vertex SSD. It seems to be somewhat cheaper.

Mar. 20, 2009

item.89246

Larry Nelson

Anandtech has a series of articles explaining how Solid State Drives work, along with their advantages and disadvantages. I recommend reading this before buying an SSD. Summary: most of them are great when writing or reading large blocks of data, but slow to a crawl when doing multiple small reads and writes.
  http://www.anandtech.com/storage/showdoc.aspx?i=3531

item.89252

Dafydd Williams

I've not heard good things about Patriot. Cheap and nasty is the general word I've had, but this could be slanderous.

AnandTech have done some work on SSD benchmarking. Their latest results are found at
http://anandtech.com/storage/showdoc.aspx?i=3531

Mar. 21, 2009

item.89302

Bob Murphy

The AnandTech article is absolutely fascinating, but also very technical and fairly dense. If you don't want to wade through it, here are some of my takeaways.

The first takeaway was around a problem called "stuttering". An SSDs that uses a cheap controller can report fantastic benchmarks and excel at things like streaming movies. But it will slow or freeze the whole computer when you use it for many mundane, everyday tasks.

How bad is stuttering? Think about noticeable pauses when typing messages into an IM window, and as reported in various NewEgg reviews, "surfing the web [is]... impossibly slow" and "[my] computer crawls while using Outlook 2007".

Fortunately, many SSDs use controllers don't cause stuttering. These include SSDs from Samsung (used in the MacBook Air), Intel, and some recent OCZ drives like the Vertex (but not older ones).

The second takeaway is that compared to other operations, SSDs and regular hard drives both suck at writing small amounts of data to random locations on the drive. Unlike regular hard drives, SSDs get worse at it as they fill up. But even slowed down that way, the better SSDs are still faster at random small writes than even the fastest regular hard drives, and random reads are unaffected by the drive filling up. (There's also a new technology called TRIM which may ameliorate this when people finally deploy it.)

One thing the AnandTech article doesn't cover is the expected usable lifespan of an SSD. SSDs use flash memory, and a flash memory block can only be erased and rewritten a limited number of times before it dies. Controllers delay failure with techniques like wear leveling, but at some point, the SSD will just wear out.

As a software developer, I beat the tar out of hard drives, so when I read in the AnandTech article that SSD blocks are only good for 10,000 writes, I was thinking of a lifespan in months. Then I read a PDF article on the Corsair website that put things in perspective. It pointed out that to wear out a Corsair 8 GB flash drive with the same 10,000-write lifetime, you'd have to write 21 GB of data to it every day for ten years. That'll work for me.

item.89298

Jeff Bagby

I want to replace the 64 GB PATA SSD drive in a 1st Gen MacBook Air with something larger. All of the larger capacity 1.8" SSD drives I can find are SATA/ZIF. Has anyone found and/or installed a larger SSD drive?

Mar. 23, 2009

item.89341

Doug Edwards

Another useful review of a OCZ Vertex SSD (128GB) is at http://benchmarkreviews.com/index.php?option=com_content&task=view&id=299&Itemid=60, (also see http://benchmarkreviews.com/index.php?option=com_content&task=view&id=270&Itemid=38)

item.89354

Dafydd Williams

I'm guessing, then, that the mid-range OCZ Core v2 series of SSDs must use a good controller. In the two Intel Macs I've got this model of SSD in, I have seen nothing but dramatic improvements across the board.

The other point to note here is that stuttering is heavily affected by file system. NTFS suffers very badly from it, but can be tweaked at format time, with some low level tools, to improve the situation. I've heard nothing about whether HFS+ is good or bad, but my own experiences seem to suggest that it runs very well indeed.

item.89362

Randall Voth

Jeff Bagby asks about replacing the drive in his original model MacBook Air. I was looking at buying a closeout for a great price but was stopped cold because of the computer's PATA drive interface. It's hard to find those drives anymore -- everything is SATA. I recently replaced a Toshiba R100 with a 2006 MacBook Pro for this reason. Drives die, it's an unavoidable fact, even for otherwise perfect computers like my beloved Toshiba (other than running Windows, of course ;-)

But the killer problem with the original MacBook Air is the ostensible drive size limitation as documented on MacOSXHints:

http://forums.macosxhints.com/showthread.php?t=95304

Any drive larger than 80gb is apparently not recognized.

Mar. 24, 2009

item.89393

Adam Glick

This should be a mandatory read regarding the state of SSD right now:
http://www.anandtech.com/printarticle.aspx?i=3531

Definitely adjusted my view of SSD.

Apr. 8, 2009

item.90196

Charles Stevenson

OCZ announces Vertex SSD for Macs:

OCZ Vertex Series Mac Edition SATA II 2.5" SSD
	Available in 30GB, 60GB, 120GB, 250GB capacities*       
	Compatible with the Apple MacBook product range       
	Onboard Cache: 64MB       
	Seek Time: <.1ms       
	Slim 2.5" Design       
	99.8 x 69.63 x 9.3mm       
	Lightweight 77g       
	Operating Temp: -10C ~ +70C       
	Storage Temp: -55C ~ +140C       
	Low Power Consumption: 2W in operation, .5W in stand by       
	Shock Resistant 1500G       
	RAID Support       
	MTBF 1.5 million hours       
	2 year warranty       
	Apple MacBook Validation List

30GB Max Performance**       
	Read: Up to 220 MB/s       
	Write: Up to 125MB/s       
	Sustained Write: Up to 80MB/s

60GB Max Performance**       
	Read: Up to 220 MB/s       
	Write: Up to 125MB/s       
	Sustained Write: Up to 70MB/s

120GB Max Performance**       
	Read: Up to 240 MB/s       
	Write: Up to 170MB/s       
	Sustained Write: Up to 100MB/s

250GB Max Performance**       
	Read: Up to 240 MB/s       
	Write: Up to 150MB/s       
	Sustained Write: Up to 100MB/s

Part Numbers       
	30GB -
OCZSSD2-1VTXA30G       
	60GB -
OCZSSD2-1VTXA60G       
	120GB- OCZSSD2-1VTXA120G       
	250GB - OCZSSD2-1VTXA250G       

The OCZ Vertex Series Mac Edition is the result of the latest breakthroughs in SSD technology. The Mac Edition delivers the incredible performance enthusiasts have come to expect from OCZ?s premium Solid State Drive offerings with the peace of mind knowing they are tested in Apple?s own labs. Additionally, the drives feature speeds up to blazing 240MB/s sequential read and 170MB/s sequential write and 64MB onboard cache, while delivering extended battery life, increased system responsiveness, and superior durability compared to conventional hard drives.

All Vertex Series SSD drives come backed a two year warranty and OCZ?s legendary service and support. Vertex series SSD drives are available in ample storage capacities of 30GB, 60GB, 120GB, and 250GB.

IMPORTANT NOTE: Solid State Drives DO NOT require defragmentation. It may decrease the lifespan of the drive.

Apr. 9, 2009

item.90246

Brad Miller

Note the [OCZ Vertex] Mac edition is the same as the regular edition, they've just cleared it through apple's labs now. Anandtech has been running a series of articles on SSD MacInTouch readers might find interesting; the most recent of which is here:

http://www.anandtech.com/storage/showdoc.aspx?i=3535

Apr. 10, 2009

item.90300

Charles Stevenson

Here come the SSD drives...

Ultra-fast 256GB Corsair SSD due out soon

[Corsair Ultra Fast 256GB SSD Sneak Peek]

We haven't put the Corsair P256 through an entire battery of tests just yet, but thought you'd like to see some preliminary benchmark data to get your geek on. According to HD Tach, the Corsair P256 offers average read speeds just over the 200MB/s mark, with average write speeds of 169.1MB/s, .1ms random access, and a 248.2MB/s burst speed. Those are pretty impressive scores that compare very favorably with Intel's excellent X25, which falls victim to the Corsair drive in average write and burst speeds.

Expected pricing for the Corsair P256 is $749. We'll have a full evaluation of the drive posted soon, complete with all pertinent technical details, so stay tuned.

Update: Corsair sent word that drives will be shipping out to distributors this week and consumers should be able to purchase then by early next week.

item.90302

Tom Francais

As mentioned in this topic, AnandTech's SSD articles are excellent. Highly recommended. Start here: The SSD Anthology: Understanding SSDs and New Drives from OCZ

Apr. 29, 2009

item.91301

Charles Stevenson

SSD prices falling as SSD goes mainstream with more products and competition.

Intel Cuts SSD Prices, Makes More 1.8-Inch SSDs for Netbooks and Small Laptops

Intel is lowering prices on several of its SSDs in a response to increased competition in the burgeoning SSD market. The world's largest semiconductor company makes SSDs using NAND flash produced by IM Flash Technologies, an Intel joint venture with Micron Technologies. [...] Increased production and competition has forced a dramatic drop in prices. Intel launched its 160GB X25-M drive for $945 just four short months ago. Today, it is available for two-thirds of its original price, a savings of over $300.
May. 7, 2009

item.91668

Robert Mohns

Jason O'Grady has published benchmarks of the Runcore Pro IV SSD drive. It's FAST. It outpaces the 5400RPM factory drive in the MacBook Pro by nearly double in sequential read/write operations, by triple in random writes, and by *six times* in random reads. That's hugely important in routine Mac tasks such as starting applications with all their little components scattered around disk.

32 GB Runcore Pro IV drives will be just $150, and 64 GB models a still reasonable $250. Bigger models get pricier of course -- 128 GB for $450, 256 GB for $890.

If you need speed more than space, that 64 GB model sure looks like a very cost effective upgrade to the drive Apple put in your MacBook or MacBook Pro...

May. 21, 2009

item.92569

Charles Stevenson

Can someone reply with performance comparisons for the new SSD drives:

Corsair P256

OCZ Summit 256GB

OCZ Vertex Mac Edition 256GB

Runcore Pro IV 256GB

G.SKILL FALCON FM-25S2S-256GBF1 2.5" 256GB SATA II MLC Internal Solid state disk (SSD)

Cavalry Pelican Elite CASD0256P2 2.5" 256GB SATA II MLC Internal Solid state disk (SSD)

Patriot Extreme Performance (EP) Warp PE256GS25SSDR 2.5" 256GB SATA II Internal Solid state disk (SSD)

FileMate 3FMS2S256M-R 2.5" 256GB SATA II MLC Internal Solid state disk (SSD)

SUPER TALENT UltraDrive ME FTM56GX25H 2.5" 256GB SATA II MLC Internal Solid state disk (SSD)

Jul. 21, 2009

item.96228

Charles Stevenson

Here are the latest prices, with manufacturer specs for read/write speeds, for currently available 2 1/2" 256 GB SSD drives:

Crucial M225 256 GB 250/200 Crucial $599
Super Talent MasterDrive SX 256 GB 220/200 NewEgg $629
Corsair P256 256 GB 220/200 NewEgg $659
Patriot Torqx 256 GB 260/180 NewEgg $699

Corsair Extreme 256 GB - not yet available
OCZ Vertex Turbo 256 GB - not yet available

item.96229

Rich Cruse

My biggest fear about having an SSD for my main drive is that it could be erased or "hosed" in a "flash"! Drives with spinning disks are bound to fail, but if the data on the drive must be recovered, it is possible to do so. I don't know if it is to recover data from a hosed SSD. Something to consider with mission critical data and computers.

Anyone?

Jul. 22, 2009

item.96308

Jeff Berg

Rich Cruse writes:

"[If] the data on the [hard disk] drive must be recovered, it is possible to do so. I don't know if it is to recover data from a hosed SSD."

Recovery from the "spinning platter" is possible but certainly not guaranteed. In either case, your first line of defense is, of course, a good backup system. (Note I said system - not a singular backup). If you need to go for data recovery, you've already done something wrong.

Jul. 23, 2009

item.96337

Rich Cruse

Most all of us know that a good back-up strategy is crucial. A lot of people know this, but never do anything about it. My point/question is: hard drives with platters "can" have data recovery performed on failed mechanisms (at great expense), but what about SSDs? Is recovery possible at all?

I think it is important to fully understand the benefits and pitfalls of SSDs before making the switch.

Perhaps Robert or someone else can explain the differences.

item.96399

Sterett Prevost

In the discussion of SSDs, let's not forget what Anandtech discovered in their in-depth SSD evaluations, some of which can be found at
  http://www.anandtech.com/storage/

What I picked up on was the limited number of writes the technology allows, the slow downs created by the convoluted way an SSD uses to save data when it's partially full or trying to clear data marked for deletion, and the need for wear-leveling routines in the firmware. I'm not ready to commit mission-critical data to them just yet, but as a startup drive that is only used for startup, an SSD would certainly enhance the speed of startup and application launch. I fondly remember the Mac Classic OS days when you could create a RAM disk with the memory control panel and use it for browser cache to speed up web browsing.

Jul. 24, 2009

item.96372

MacInTouch Reader

Intel Delivers Industry's First 34-Nanometer NAND Flash Solid-State Drives; Advancement Lowers Prices by Up to 60 Percent [press release]

Intel Corporation is moving to a more advanced, 34-nanometer (nm) manufacturing process for its leading NAND flash-based Solid State Drive (SSD) products, which are an alternative to a computer's hard drive. The move to 34nm will help lower prices of the SSDs up to 60 percent for PC and laptop makers and consumers who buy them due to the reduced die size and advanced engineering design.

The multi-level cell (MLC) Intel(R) X25-M Mainstream SATA SSD is aimed at laptop and desktop PCs and available in 80 Gigabyte (GB) and 160GB versions. SSDs are data storage devices found inside computers. Because SSDs have no moving parts they offer faster performance and greater energy efficiency and durability than traditional hard disk drives (HDDs). A draw for gamers, media creators and technology enthusiasts, SSDs have also played a key role in the emergence of ultra-thin and light notebook PCs that are becoming increasingly popular due to their design, size and longer battery life.

"Our goal was to not only be first to achieve 34nm NAND flash memory lithography, but to do so with the same or better performance than our 50nm version," said Randy Wilhelm, Intel vice president and general manager, Intel NAND Solutions Group. "We made quite an impact with our breakthrough SSDs last year, and by delivering the same or even better performance with today's new products, our customers, both consumers and manufacturers, can now enjoy them at a fraction of the cost."

The Intel X25-M on 34nm flash memory is drop-in compatible with the current 50nm version and will continue to be drop-in compatible to replace existing hard disk drives (HDDs).

Compared to its previous 50nm version, the new Intel X25-M offers improved latency and faster random write Input/Output Operations Per Second (IOPS). Specifically, Intel's new SSD provides a 25 percent reduction in latency, for quicker access to data, operating at 65-microsecond latency compared to approximately 4,000 microseconds for an HDD.

Random write performance increases twofold, further separating the X25-M from other competing SSDs. By delivering up to 6,600 4KB write IOPS and up to 35,000 read IOPS, the X25-M continues to set the bar for SSDs, while leapfrogging HDDs which only operate at several hundred IOPS. This provides for markedly faster system and application responsiveness. These improvements in latency and IOPS not only benefit desktop and notebook users, but also server and workstation users, as they utilize Intel's cost-effective, yet performance-oriented, MLC SSDs for enterprise computing.

New channel prices for the X25-M 80GB are $225 for quantities up to 1,000 units (a 60 percent reduction from the original introduction price of $595 a year ago). The 160GB version is $440 (down from $945 at introduction) for quantities up to 1,000 units. The X25-M comes in a standard 2.5-inch form factor. The X18-M, in a 1.8-inch form factor, will begin shipping on 34nm later in the quarter.

Drop-in compatible with SATA-based HDDs and all operating systems, the X25-M will also support Microsoft Windows 7 when it becomes available. At that time, Intel plans to deliver a firmware update to allow support of the Windows 7 Trim command, along with an end user tool, to allow users to optimize the performance of their SSD on Windows XP and Vista operating systems.

For further information on Intel's High Performance line of solid-state drives, including the Intel X25-M Mainstream and the Intel(R) X25-E Extreme SATA SSD targeted for server, workstation and enterprise storage applications, visit www.intel.com/go/ssd.

item.96446

Skot Nelson

Re:

I'm not ready to commit mission-critical data to them just yet

At work we use an IBM AIX box to run our ERP on the (strangely disconcerting) IBM Universe system.

We copy this database to another box which is used for our data analysis tool (MITS Report & MITS Discover.) This copied database is entirely hosted on an SSD RAID.

In 1.5 years there've been no problems with the SSDs themselves. We did do a firmware update, but the drives weren't actually exhibiting problems.

Arguably this is not a "mission-critical" application: it's a copy of our mission critical data, but the data analysis tools are used daily throughout the company.

I don't think they're economically worth it for mainstream applications. In this case the MITS Discover cube building process on the Universe multi-value database was taking so long that the plunge was taken before I arrived. It apparently vastly improved the read/write times, and reduced a several hour process a day to about 20 minutes.

I might be ready to trust mission critical data to SSDs, but that economic issue is very real.

Jul. 25, 2009

item.96476

Mike Viksna

Here's my question regarding SSD's write limitation and OS X: The automagic defrag that OS X does behind the scenes obviously writes data. Does this process not occur when an SSD is the boot drive? A boot drive is subject to a bit more change than an attached drive used for storage and retrieval. Lots of log files and databases being opened, updated and closed. Just curious when those early MBAir's may start exhibiting senior moments? And how will we know if and when bit rot is setting in?

Jul. 27, 2009

item.96521

Dave Lockwood

From: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3605, 21st July:

"A little known fact about the original X25-M was that its controller wasn't Halogen-free. Because Intel used Halogens in the first controller, companies that had strict environmental restrictions (e.g. Apple) wouldn't touch the drives"

Maybe I'm getting old and cynical, but I'd bet if Apple *had* been able to offer these drives BTO, there'd never have been a firmware problem with eSATA speeds ;)

Jul. 28, 2009

item.96578

Lyman Taylor

Apple's position is that the classic defragmentation tools are most often not necessary.

http://support.apple.com/kb/HT1375

So the "automatic", default fragmentation is to fix 'hot' files. If limited to just the log files and reasonable sized files then shouldn't be much of a problem. (the autodefrag avoids files larger than 20MB)

(also earlier: http://www.macintouch.com/readerreports/macosx10_3_8/topic2779.html)

Apple's points on larger drives is still just as relevant for SSDs. You are going to be in a different circumstances if you fill your SSD to 75+% capacity than if fill it to 30-50% capacity. If there is more free space then the wear leveling is going to get more spread out (with Apple's strategy layered on top of the SSD's internal one). [ To some extent the strategies are similar in that SSDs have more then their rated amount of storage space. The internal strategy is to use the "extra' space to even out the 'wear'. ]

Moving every single file around so that eliminate fragmentation almost completely isn't really necessary if you are smarter about read ahead cache and aren't looking for doing peak load, large database table scans.

SSDs do wear out, but I think it is a more complicated set of factors ( space used and rewriting done ) then it is with spinning drives ( times turned on/off and long periods of random access).


P.S. Apple moving hot, read-only files to the "faster" part of the spinning drive are probably useless on a SSD but shouldn't hurt much if only done occassionally.


Jul. 29, 2009

item.96642

Kynan Shook

Regarding SSDs, OS X auto defragmentation, and wear leveling: it shouldn't be much of a concern. I don't know the specs for the drives Apple uses, but most SSDs are good for a few million writes before they become unreliable. Say you have am expectation of using your drive for 5 years before replacing it (or the computer). Divide by 1 million - that gives you 158 seconds. So, you'd have to write all the empty space on your drive (including the parts reserved for wear leveling) evey 158 seconds to hit that write cycle specification, assuming a perfect wear leveling algorithm that only uses free space. Even with a fairly small amount of free space (say a gig or so), you're not likely to come anywhere close to that, even if the OS remains completely oblivious to the fact that it's running on an SSD.

Aug. 4, 2009

item.96969

Geoff Strickler

Regarding Kynan Shook's July 29 comments on SSDs:

While SLC (e.g. Intel X25-E) flash can support at least 100K write cycles and typically 1M or more, the mass market SSDs are based upon MLC flash because of its higher density and lower cost. MLC flash is only rated for 10k write cycles. That's still plenty for most consumer applications, especially when you consider the effects of write combining and wear leveling features, however, it's nowhere near the 1M cycles you may see from an "enterprise" class SLC flash SSD.

The spec sheet for the Intel X25-M products states:

"The drive will have a minimum of 5 years of useful life under typical client workloads with up to 20GB of host writes per day".

Mar. 10, 2010

item.110639

Scott Newman

The OCZ Vertex Limited Edition SATA SSD has the same Sandforce controller as the new SSD's from Other World Computing. They are 2.5" SATA compatible.

Yet, their Web site [here] says

"Compatible with Windows XP, Vista, 7, and Linux."

This is absurd and simply reinforces the prejudice that many Windows people have against Macs.

item.110736

Robert Mohns

Scott Newman wrote:

"The OCZ Vertex Limited Edition SATA SSD has the same Sandforce controller as the new SSD's from Other World Computing. They are 2.5" SATA compatible.
Yet, their Web site [here] says

"Compatible with Windows XP, Vista, 7, and Linux."

This is absurd and simply reinforces the prejudice that many Windows people have against Macs."

While it probably does reinforce that prejudice, I don't think it's absurd. The simple explanation is they'd like to avoid having to provide technical support to the Mac platform.

Consider that Apple's latest iMacs can require different hard drive cables depending on which manufacturer's drive you buy. I can't really blame anyone for not wanting to support products made by a company that does that! (Remember what a pain Compaqs were 20 years ago?)

Supporting Apple hardware can be very annoying for third party vendors. Ask any vendor who ended up sitting on unsellable "Bondi Blue" accessories after the iMac's went all rainbow-colored. :-)

That said, I think OCZ is passing up a great opportunity to sell to performance-oriented Mac users. Mac users are already used to paying a premium for better quality, so why not sell their fast, pricey hardware to us? Their loss... and Otherworld Computing's gain!

item.110752

MacInTouch Reader

Scott Newman writes:

"The OCZ Vertex Limited Edition SATA SSD has the same Sandforce controller as the new SSD's from Other World Computing. They are 2.5" SATA compatible.

Yet, their Web site [here] says

'Compatible with Windows XP, Vista, 7, and Linux.'

This is absurd and simply reinforces the prejudice that many Windows people have against Macs."

Or it just reinforces that not everyone can afford Apple hardware to test and verify the device. Or wish to add supporting Macs to their customer care services. Because even though it should 'just work', if it doesn't, and you call them, you won't accept "Well, it works for us on our one Mac!" as an excuse, you'd complain that they claimed Mac compatibility but not offer it.

OWC, on the other hand, is a Mac shop and so one would expect that the products they sell would include Mac support.

item.110758

Brad Riendeau

The web site for the OCZ Vertex Limited Edition uses the classic and basically meaningless buzz words that follow

"...cutting-edge new ...the epitome of innovation for power users seeking to integrate a new class ... For early adopters, ... can take a high-end computing experience to the next level."

Could be beer, could be a video game, could be sports equipment but the experience will be high end you will be powerful in a new class and moving to the next level. They don't list Macs because ... maybe we're all ready there.

item.110772

Stu MacKenzie

Re: Scott Newman's comment on Vertex Limited SSDs for Win and linux

It was my impression that OCZ had trimmed the speeds of its Vertex series for Mac SSDs owing to Apple's "journaling" driver. My Google-fu is weak today; the only mention of this I could quickly turn up was this article from a year ago over at BSN. (Maybe this Anandtech article goes into it, but...tl;dr ;-)

If SATA driver issues are the straight story, I appreciate OCZ throttling the drives in exchange for solid ("certified") performance.

Mar. 11, 2010

item.110806

David Simpson

There is a note from a customer on the NewEgg website [here] regarding this issue with Macs:

Fast but Flawed
Reviewed By: teknishn on 3/3/2010
Tech Level: high - Ownership: less than 1 day
This user purchased this item from Newegg
Pros: Unbelievably fast. Fastest overall drive money can buy right now.
Cons: Has a nasty firmware bug that causes the drive to brick on Macs after OSX goes to sleep.
Other Thoughts: Wait for new firmware if you own a Mac

item.110855

Wil Nelson

This bricking after the computer goes to sleep can be fixed by going to System Preferences>Energy Saver and unchecking "Put hard disk to sleep when possible" both in the Battery and Power Adapter panes.

Mar. 18, 2010

item.111192

Wil Nelson

In addition to restarting using the power button and setting/resetting the above indicated Energy Preferences if one is using Smart Sleep it should be set to Hibernate.

Apr. 26, 2010

item.113405

Lee Clawson

Regarding Robert Mohns benchmark test of a MacBook Air, 2.13 GHz, 120 GB SSD...

I've been told that Snow Leopard does not support the "trim" command. If so, should I be concerned? And will it affect performance of a SSD suffer over the long run?

item.113410

Robert Mohns

Lee Clawson asks:

"I've been told that Snow Leopard does not support the "trim" command. If so, should I be concerned?? And will it affect performance of a SSD suffer over the long run??"

As of 10.6.3 Snow Leopard does not support TRIM (which not all SSDs support, just newer firmware). First, some basics for those who don't know (yet!) what we're talking about...

TRIM is a new firmware command added to performance-oriented solid state drives that erases unused drive blocks. And why is that important? Because flash memory must erase a block before writing data to it (details at Wikipedia: Flash memory - block erasure).

It may help to think of CD-RW (rewritable), which can be written at full speed, but must be erased before you can overwrite existing data. That erase takes a while, but once done, you're back to full speed writing.

From the reading I've done, we're talking about a 20-30% reduction in write speeds when a block is overwritten. (The SSD first must erase the block, then write the data.)

While performance will "suffer", an SSD is still so much more responsive than a hard drive -- hardly any latency at all -- that it will still be a huge net win. The vast majority of drive activity on a Mac is reads: starting apps, reading preference files, Spotlight database lookups, etc. Writes aren't much of an issue for most users compared to reads. Engadget wrote in one report "the difference between fast and really, really fast is not as distinguishable as we like to think."

I wish Apple would add TRIM support, but I understand why they haven't yet: It's a bit of a moving target at the moment, with implementations evolving pretty quickly, and there have been some serious SSD bugs already. (One version of Intel's fast SSD's corrupted data if a BIOS password was set and then changed. Ooops.)

The first law of writing disk drivers and file managers is "Don't Lose Data" (and TRIM involves both parts cooperating). Apple has run screwed this up a few times over the years -- remember the catastrophic 10.5 file move bug? Previous to that, there was a serious issue with Panther Firewire and the popular Oxford 922 firewire chipset resulting in corrupted data. And back in 2002, an earlier version of the file move bug appeared in Mac OS X 10.1.3.

With this track record, I would really prefer that Apple play it safe, even if not delivering the maximum possible performance. My data is too important to risk to a half-baked driver that was rushed to market.

At least one current SSD has added "garbage collection" to automatically erase unused blocks during idle time, getting them ready for fast writes. However, this only works with Windows NTFS and FAT file systems, so Mac users won't see that benefit.

Without TRIM, we'll see different (slower) benchmarks for a new drive than we'd see for a drive that's been filled up. But as both I and Engadget noted, the actual user experience difference isn't very noticeable, since most of your Mac's day-to-day drive use is reading, not writing.

Apr. 27, 2010

item.113488

MacInTouch Reader

Robert Mohns wrote yesterday:

At least one current SSD has added "garbage collection" to automatically erase unused blocks during idle time, getting them ready for fast writes. However, this only works with Windows NTFS and FAT file systems, so Mac users won't see that benefit.

I believe OCZ Technology's implementation of garbage collection works with Macs (and other non-Windows computers), not requiring NTFS or FAT to maintain high performance over time. (I don't know why any manufacturer's GC algorithm would require NTFS or FAT). OCZ has sold special Mac versions of their Vertex SSDs at a premium price and slightly reduced peak performance, but I don't know that the Mac versions are any longer necessary, what with more recent Mac OS X and OCZ firmware updates. I have one Mac version and one non-Mac version of the Vertex 250GB drive and have had zero problems with them for many months now, as well as great long-term performance.

One of the 250GB SSDs is used in a RAID1 mirror managed by a Mac Pro RAID card. The second disk in the mirrored set is a traditional platter drive that happens to be 1TB 7200RPM (750 GB wasted). The hybrid disk technologies affords better protection than using SSD alone, with overall performance on par with SSD. Activity from the HDD is rarely heard at all, since the SSD responds faster to nearly all requests and most requests are read requests. It's basically like running off an SSD with a live HDD backup silently available in case of trouble.

Just one word of warning: using SSD in a fast computer can be very addictive!

item.113491

Lee Clawson

Robert Mohns writes:

But as both I and Engadget noted, the actual user experience difference isn't very noticeable, since most of your Mac's day-to-day drive use is reading, not writing.

Thanks for a very complete and informative answer.
One other question comes to mind; what about the performance of an SSD with graphic applications (especially image processing apps) that make use of a "scratch disk" where I would expect lots more (temporary) writing to disk ???

item.113497

David Foster

I've been using a OCZ 250GB flash drive in both a MacBook Pro and a Mac Pro (early 2009) and I am very pleased with the results. Although I haven't really detected a slow-down yet, I was wondering what is the best recommended restorative method for rejuvenating these drives if I do eventually see a discernible drop in performance. Would you recommend cloning to a fast HD, reformatting, and then cloning the data back using an utility like SuperDuper? One more question: Drive Genius 3 has been released and it has a hard drive monitoring feature (I think it's more than just a SMART monitor). Could this type of feature (which I would think continuously probes drive action in idle-time) be harmful (or at least not beneficial) to a SSD boot drive? I will query the vendor and relay their response back here.

Apr. 28, 2010

item.113546

Lyman Taylor

Re:

I believe OCZ Technology's implementation of garbage collection works with Macs (and other non-Windows computers)

There are primarily two pools to collect garbage from. One is from blocks that have been "abandoned" by the file system. The second is from blocks "abandoned" by the wear leveling system. The first is going to require being hooked into the file system some way since its metadata drives what is "freed"/"abandoned". The second taps into the wear leveling, logical/virtual mapping metadata that the drive keeps track of.

The ones at the file system level are going to be nice for recovering larger collections and perhaps removing fragmentation. However, when leveraged so that effectively have a "TRIM" (collection of freed blocks) data recovery tools maybe significantly impeded if file system metadata is also collected aggressively.

The upside for the drive local metadata collection is that it won't have that secondary effect. (it is no worse than the current hard drives which overwrite blocks by default. Previous versions of metadata blocks that are updated aren't accessible in either (minus extraordinary surface scans)). In classic garbage collection, this is like the short ephemeral collectors.

The upside on the file system metadata collection is that it will briefly throttle down data requests while doing this since requests coming down same pipeline. It is also cheaper for the drive folks since the "smarts" are in the computer, not the internal drive controller.

Re:

since the SSD responds faster to nearly all requests and most requests are read requests. It's basically like running off an SSD with a live HDD backup silently available in case of trouble.

Which drives do you do the backups from then? In other words, do you periodically pull the HDD and put in a new one to do a back up or do you run a backup/clone program to replicate data for this logical RAID 1 drive ? The problem is that if almost never read from the HDD you won't know if the HDD has a bitrot problem. If you have grossly unbalanced reads in a RAID 1 system and not checksuming the data from the difference sources against each other (unlikely if not running something like ZFS or Netapp which tag all the blocks with checksums so that is a simple 'compare'.), then not going get signals in some cases when that HDD goes wonky.

If the RAID card goes out periodically proactively patrolling for bit rot (comparing checksums across redundant data) then not so much a problem. The system would be unbalanced read most of the time, but periodically would look for the problem. If not and has this "only talk to fastest" heuristic they going to get skewed view of the data that can drift over a extended period of time.

Typically, HDD goes bad in writes mode as much as read if it is some mechanical problem. So it is not a huge increase in risk, but it is present.

Additionally, most SSD drives are a bit kneecapped in speed so at full blast streaming, a RAID card will ask HDD to get some blocks because the SSD is maxed out. However, as SSD and HDD are more extremely different the problem manifests itself more, because the imbalance will be greater.

item.113536

Lyman Taylor

Re:

Would you recommend cloning to a fast HD, reformatting, and then cloning the data back

Unless you are using a specialized reformat utility program specific to your SSD, this can have the opposite effect.

The cheap way to reformat a drive is to go to the metadata catalog and mark all of the blocks as "blank"/"unused". The drive itself is not 'informed' that the blocks are marked that way because the metadata is itself just data. This is why drive data recovery programs work. The data was never erased, just abandoned. If a low level program wanders around over all the blocks, it can "find" most of it again if willing to piece back together the jigsaw puzzle (and haven't lost too many critical pieces to the puzzle.)

The drive vendor will have something that goes block by block marking them as unused. This will bypass the logical/virtual block interface the wear leveling subsystem presents and tweak the metadata the leveling system uses. (Similar issue for "zeroing" data blocks. Normally not happening on SSD because no access to the physical blocks. The writes aren't necessarily to the same physical location, just the same logical (virtual) one.)

In short, there are multiple layers of metadata on a SSD. One for the outside view of the data and one for the inside view of the data. SSDs can slow down when those two get severely out of sync.

item.113558

Stephan Wik

I just installed a 100GB OWC Mercury Extreme Enterprise SSD in my MacBook Pro and I am very happy with it. It is now the fastest, quietest and coolest laptop I've ever used, and the run time on a full battery charge has been greatly extended.

Mail and Safari now both launch with one 'bounce' i.e. very fast indeed. I found that iDefrag 2.0 did an excellent job in getting the SSD into good shape after I'd restored from my Time Machine backup. Booting and application launching definitely speeded it up significantly after a full defrag(i.e launching Safari went from 2 bounces to 1).

Standard disclaimer: no connection to either company, just a happy customer.

Next Page...


MacInTouch Amazon link...

Talk to MacInTouch     Support  •  Find/Go